Development of sustainable precision farming systems for swine: Estimating realtime individual amino acid requirements in growing-finishing pigs

Abstract

The objective of this study was to develop and evaluate a mathematical model used to estimate the daily amino acid requirements of individual growing-finishing pigs. The model includes empirical and mechanistic model components. The empirical component estimates daily feed intake (DFI), BW, and daily gain (DG) based on individual pig information collected in real time. Based on DFI, BW, and DG estimates, the mechanistic component uses classic factorial equations to estimate the optimal concentration of amino acids that must be offered to each pig to meet its requirements. The model was evaluated with data from a study that investigated the effect of feeding pigs with a 3-phase or daily multiphase system. The DFI and BW values measured in this study were compared with those estimated by the empirical component of the model. The coherence of the values estimated by the mechanistic component was evaluated by analyzing if it followed a normal pattern of requirements. Lastly, the proposed model was evaluated by comparing its estimates with those generated by the existing growth model (InraPorc). The precision of the proposed model and InraPorc in estimating DFI and BW was evaluated through the mean absolute error. The empirical component results indicated that the DFI and BW trajectories of individual pigs fed ad libitum could be predicted 1 d (DFI) or 7 d (BW) ahead with the average mean absolute error of 12.45 and 1.85%, respectively. The average mean absolute error obtained with the InraPorc for the average individual of the population was 14.72% for DFI and 5.38% for BW. Major differences were observed when estimates from InraPorc were compared with individual observations. The proposed model, however, was effective in tracking the change in DFI and BW for each individual pig. The mechanistic model component estimated the optimal standardized ileal digestible Lys to NE ratio with reasonable between animal (average CV = 7%) and overtime (average CV = 14%) variation. Thus, the amino acid requirements estimated by model are animal- and time-dependent and follow, in real time, the individual DFI and BW growth patterns. The proposed model can follow the average feed intake and feed weight trajectory of each individual pig in real time with good accuracy. Based on these trajectories and using classical factorial equations, the model makes it possible to estimate dynamically the AA requirements of each animal, taking into account the intake and growth changes of the animal. © 2012 American Society of Animal Science. All rights reserved.